benchgap
Other · model

Agents-A1-4B benchmark scores

As of 2026-11-22, Agents-A1-4B (Other) has measured scores on 8 benchmarks and estimated scores on 39 more.

BenchmarkScoreSource
AA-SciCode30.5%estimated ± 3.0 pp, high confidence
GDPval-AA26.2%estimated ± 7.8 pp, medium confidence
IFBench69.1%measured
AA-LCR55.7%estimated ± 12.5 pp, low confidence
LiveCodeBench v659.6%measured
SciCode29.6%measured
BrowseComp66.8%measured
HLE w/ tools29.5%estimated ± 5.5 pp, low confidence
IFEval94.8%measured
LongBench v252.1%measured
VITA-Bench40.3%measured
τ²-bench results78.2%measured
AA Coding Index11.8%estimated ± 5.7 pp, medium confidence
APEX-Agents-AA14.3%estimated ± 4.4 pp, high confidence
DeepSearchQA78.8%estimated ± 6.8 pp, low confidence
SWE-bench Verified47.5%estimated ± 5.6 pp, medium confidence
BFCL v460.4%estimated ± 8.5 pp, low confidence
CyberGym53.9%estimated ± 7.3 pp, medium confidence
JobBench25.2%estimated ± 10.7 pp, low confidence
SkillsBench38.6%estimated ± 4.1 pp, low confidence
SWE-bench Pro0.0%estimated ± 6.0 pp, low confidence
WideResearch70.9%estimated ± 3.5 pp, high confidence
AA Agentic Index22.0%estimated ± 8.2 pp, medium confidence
AA-IFBench56.7%estimated ± 7.4 pp, medium confidence
Vals LiveCodeBench81.3%estimated ± 0.8 pp, low confidence
Gert Labs30.9%estimated ± 11.1 pp, low confidence
AA Harvey LAB v1.087.2%estimated ± 9.8 pp, low confidence
AA Tau3 Banking26.8%estimated ± 3.6 pp, low confidence
MazeBench0.0%estimated ± 6.4 pp, low confidence
OSWorld-Verified59.4%estimated ± 5.7 pp, medium confidence
Vals SWE-bench31.6%estimated ± 5.3 pp, low confidence
CWE-Bench-AA44.1%estimated ± 6.4 pp, medium confidence
AA Cyber Index33.2%estimated ± 13.5 pp, low confidence
DeepsecBench-AA4.9%estimated ± 9.0 pp, low confidence
AA ITBench39.7%estimated ± 4.1 pp, high confidence
DeepSWE24.3%estimated ± 8.4 pp, low confidence
DRACO65.9%estimated ± 2.9 pp, low confidence
Toolathlon-Verified51.5%estimated ± 7.4 pp, medium confidence
Claw-Eval62.7%estimated ± 6.4 pp, medium confidence
DeepPlanning26.9%estimated ± 9.1 pp, low confidence
MCP Atlas54.0%estimated ± 10.3 pp, low confidence
QwenClawBench52.0%estimated ± 2.0 pp, medium confidence
Toolathlon37.2%estimated ± 4.2 pp, high confidence
ResearchClawBench15.8%estimated ± 2.3 pp, medium confidence
Agents' Last Exam16.8%estimated ± 11.6 pp, low confidence
LifeVal51.4%estimated ± 1.6 pp, medium confidence
PinchBench89.4%estimated ± 3.6 pp, medium confidence